May, 1930.

] 919

AMERICAN SOCIETY OF CIVIL ENGINEERS

INSTITUTED 1852

PAPERS AND DISCUSSIONS

Tbis Society is not responsible for any statement made or opίnion expresβed
in its publications.

ANALYSIS OF CONTINUOUS FRAMES

ΒΥ DISTRIBUTING FIXED-END MOMENTS
ΒΥ HARDY 0Ross,* Μ. ΑΜ. Soc. Ο. Ε.

SYNOPSIS

The purpose of this paper is to e:xplain a method which has ./

been found useful in analyzing frames which are statistiM.ll indeterminate.
The essential idea which the writer wi8hes to present involves no mathematical
relation8 except the 8implest arithmetic. It is true that in order to apply the
method it is neces8ary to determine certain constants mathematically, but the
1nean8 to be used in determining these con8tant8 are not di8cussed in the
paper, nor a1·e they a part of the method. These constants have been derived
by 80 many writer8 and in 80 many slightly di:fferent ways that there is little
occasion to repeat here the whole procedure.
The reaction8 in beam8, bents, and arche8 which are immovably fixed at
their ends have been extensively di8cussed. They can be found comparatively
readily by methods which are more or less 8tandard. The method of analysis
herein presented enables one to derive from these the moments, shears, and
thru8t8 required in the design of complicated continuous frames.

DEFINITIONS

For convenience of reference, de:finitions of three terms will ,be introduced

at once. These terms are ":fixed-end mψnent", "sti:ffness", and "carry-over
factor".
By ":fixed-end moment" in a member is meant the moment which would
exist at the ends of the member if its ends were :fixed against rotation.
"Sti:ffness", as herein used, is the moment at one end of a member (which
is on unyielding supports at both ends) necessary to produce unit rotation of
that end when the other end is :fixed.
Π one end of a member which is on unyielding supports at both ends is
rotated while the other end is held fixed the ratio of the moment at the :fixed
NoτE .-Written dίscussion on this paper will bc closed in September, 1930, Proceedin-gs.
• Prof. , Structura1 Eng., Univ. of Illinols, Urbana, Ill.
 Γ

920 .ΛN.ALYSIS OF CONTINUOUS FR.AMES [Papers.

end to the moment producing rotation at the rotating end is herein called
the "carry-over factor."
EFFECT OF J ΟΙΝΤ RoTATION

Imagine any joint in a structure, the members of which a1·e being deformed
by loads, or in some other way, to be first held against rotation and then
released. Call the algebraic sum of the fixed-end moments at the joint the
"unbalanced fixed-end moment". Before the joint is release.d this unbalanced
fixed-end moment will not usually be zero; after the joint is released, the sum
of the end moments at the joint must be zero. The total change in end
moments, then, must equal the unbalanced fixed-end moment. This may be
stated in another way by saying that the unbalanced fixed-end moment has
been "distributed to" the connecting members in some ratio.
When the joint is released all connecting members rotate through the same
angle and this rotation at the end is accompanied by a change in end moment.
The change in end moments is proportional to the "sti:ffness" of the members.
Hence, it may be said that when the joint is released the unbalanced
fixed-end moment is distributed among the connecting members in proportion
to their sti:ffness.
The rotation of the joint to produce equilibrium induces moments at the
other ends of the connecting members. These are equal in each member to
the moments distributed at the rotating joint multiplied by the carry-over
factor at the ~otating end of the member. This follows from the definition
of "carry-over factor".
ΜοΜΕΝΤ DISTRIBUTION

The method of moment distribution is this: (α) Imagine all joints in the
structure held so that they cannot rotate. Compute the moments at the ends
of the members for this condition; (b) at each joint distribute the unbalanced
fixed-end moment among the connecting members in proportion to the con-
stant for each member defined as "sti:ffness"; (c) multiply the moment dis-
tributed to each member at a joint by the carry-over factor at that end of the
member and set this product at the other end of the member; (d) distribute
these moments just "carried over"; (e) repeat the process until the moments
to be carried over are small enough to be neglected; and (f) add all moments
-fixed-end moments, distributed moments, moments carried over-at each end
of each member to obtain the true moment at the end.
Το the mathematically inclined the method will appear as one of solving
a series of normal simultaneous equations by successive approximation. From
an engineering viewpoint it seems simpler and more useful to think of the
solution as if it were a physical occurrence. The beams are loaded or other-
wise distorted while the joints are held against rotation; one joint is then
allowed to rotate with accompanying distribution of the unbalanced moment
at that joint and the resulting moments are carried over to the adjacent
joints; then another joint is allowed to rotate while the others are held against
rotation; and the process is repeated until all the joints are "eased down" into
equilibrium. ·
May, 1930.] ANALYSIS OF CONTINUOUS FRAMES 921

ΒΕΑΜ ΟοΝsτΑΝτs
This method of analysis is dependent on the solution o:f three problems in
the mechanics o:f materials. These are the determination o:f the :fixed-end
moments, o:f the sti:ffness at each end, and of the carry-over :factor at each end
for each member of the frame under consider~tion. The determination of
these values is not a part of the method of moment distribution and is not
discussed in this paper.
The sti:ffness of a beam of constant section is proportional to the moment
of inertia divided by the span length, and the carry-over factor is - i.
The proof or derivation of these two statements and the derivation of
formulas for :fixed-end moments is left to the reader. They can be deduced
by the use of the calculus; by the theorems of area-moments; from relations
stated in Bulletin 108 of the Engineering Experiment Stationofthe University
of Illinois (the Slope-De:flection Bulletin); from thetheoremofthreemoments;
by what is known to some as the column analogy method ;* or by any of the
other corollaries o:f geometry as applied to a bent member. Formulas for
:fixed-end moments in beams of uniform section may be found in any struc-
tural handbook.
SroNs oF ΤΗΕ BENDING ΜοΜΕΝτs

It has seemed to the writer very important to maintain the usual and
familiar conventions for signs of bending moments, since these are the
conventions used in design.
For girders the usual convention is used, positive moment being such as
sags the beam. For vertical members the same convention is applicable as
for girders if the sheet is turned to read from the right as vertical members
on a drawing are usually read. The usual conventions for bending moments
are, then, applicable to both girders and columns if they are looked at as a
drawing is usually lettered and read.
Moments at the top o:f a column, as the column stands in the structure,
should be written above the column and those at the bottom of the column,
as the column stands in the structure, should be written below the column
when the sheet is in position to read the columns. This is necessary because
positive moment at the right end of a beam and at the top of a column both
represent tendencies to rΌtate the connected joint in the clockwise direction.
It makes no di:fference whether girder moments are written above or below
the girder. Either arrangement may be convenient. Oonfusion will be avoided
by writing column moments parallel to the column and girder moments par-
allel to the girders.
When any joint is balanced the total moment to the right and to the left
of the support is the same, both in absolute value and in sign. The unbal-
anced moment is the algebraic di:fference of the moments on the two sides of
the joint.
LIMITATION OF METHOD

From the fact that the terms, "sti:ffness?' and "ca:ι;ry-over factor", have
been defined for beams resting on unyielding supports, ' it follows that direct
• "The Column Analogy," by Hardy Cross, Μ. Α . Soc.
Exper im ent Stat ί on, Unίv . of Πlinois, Urbana, Ill. ('In press.

)( 2 /.J
922 ΑΝ ALYSIS OF CONTINUOUS FRAMES [Papers.

application of the method is restricted to those cases where the joint do not
move dur'ng the process of moment distribution. , The method, however, can
be applied in an indirect way to ,cases in which the joints are displaced during
the moment distribution, as indicated later.
As the method has been stated, it is restι·icted only by this condition that
the joints are not displaced. If this condition is satisfied it makea no differ-
ence whether the members are of constant or of varying section, curved or
straight, provided the constants, (α) fixed-end moments at each end, (b) stiff-
ness at each. end, and (c) carry-over factor at each end, are known or can be
determined. Such values can be derived by standard methods and may be
tabulated for different types of members and conditions of loading.
It will be found that i.n most cases accuracy is needed only in the fixed-end
moments. It does not ordinarily make very much difference how, within
reason, the unba]anced moments are distributed, nor, within reason, how much
of the distributed moments are carried over.
In the illustration which follows it has been assumed that the members are
straight and of uniform section. The stiffnesses, then, are proportional to the
moments o:f inertia, (Ι), divided by the lengths, (L), but the rel~tive values
Ι
given for L in this problem might quite as well be the relative stiffness of a
series o:f beams o:f varying section. In this latter case, however, the carry-over
factors for the beams would not be- i.

Β
::1[8 ~ ~ ~ ~
Oocioόo...;o....;oMo
,..:
~ -100.00 -100.' ο
7ιιιιι7 ~~:gb +62.50 -37.50
+ 6.25 + 5.50
+ 3.33 1
- 0.78 1 ... 0.47
- 5.34 + 6.68 ... 9.38
+ 3.56 ..1.69 - 1.01
- 0.55 ...
0.69 1 - 0.12
+ 0.37 - 0.51 + 030
•0.48 .. 0.25
® ~ g:~~ -0.15 1 ... 0.08
- 0.10 + 0.13 - 0.08
.::J1:QZ - 0.13 + 0.08
-37.12 --::ms-
1

Α
FIG. 1.

lLL USTRA ΤΙΟΝ

The illustration given (Fig. 1) is entirely academic. It is not intended

to represent any particular type of structure nor any, probable condition o:f
loading. It has the advantage :for the purpose o:f this paper that it involves all
the conditions that can occur in a frame which is made up o:f straight members
1
and in which the joints are not displaced.
May, 1930.] .ΛN.ALYSIS OF CONTINUOUS FR.AMES 923

The loads on the :frame are supposed to be as indicated. The relative

Ι
νa1ues of L for the different members are indicated in circles.

The fixed-e~d moments in all members aι·e fiτst written. In this problem
they are arbitrarily assumed to be as shown, as follows: at Λ, Ο; at Β, in
Β Λ, Ο, and in Β .Ο,- 100; at Ο, in Ο Β,- 100, in Ο lι\ +·
80, in Ο D,- 200,
and in Ο G, -50; at F, + 60; at G,- 50; at D, in D Ο, - 100, and in D Ε, Ο;
at Ε, in Ε D, Ο, and in the cantilever, - 10.
Be:fore proceeding to a solution o:f the problem, attention may be called
to the arraηgement o:f the computations. The moments in the girders are
wι·itten parallel to the girder·s; those in the columns, parallel to~ the columns.
The original fixed-end moments are written next to the members in which they
occur, the successive moments distributed or carried over being· written above
or below these, but farther :from the member.
The arrangement of the moments in the columns in positions above the
columns, when the paper is turned into a position to write these moments,
:for the top of the columns (at Β, F, and 0), and in positions below the col-
umns for the bottom of the columns (at Λ, Ο, and G), is an essential part of
the sign convention adopted.
The rnoment at Ο in the girder, Β Ο, is written above the giι·der in order
to get it out of the way. Otherwise, it · makes no difference whether the
moments are written above or below the girder.
The signs of the :f:ixed-end moments are determined by observing the direc-
tion of .flexure at the ends of the members due to the loads. In order to apply
to the columns the ordinaιΎ conventions for signs of bending moments it is
necessary to turn the drawing of the structure.
The reader should realize that the solution is built up step by step. It
is always the last :figures showing that are to be operated on-distributed or
carried ove:r-so that in ordinary framework there is little chance for con-
fusion as to what step should be taken next.
Distribute at each joint the unbalanced moment, as follows:
· 1.-At Λ there is no moment.
2.-At Β there is an unbalanced moment of -100 on one side of the joint.
This moment is distributed to Β Λ and to Β Ο in the ratio, 2 : 4, so that the
2 4
distributed moment to Β Α is - - - 100 = 33.33 and to Β 0, - - 100 = 66.67.
2 +4 2+4
The signs are written in the only way possible to balance the joint by
giving the same total moment (- 33.33) both to left and right of the joint.
3.-At Ο, the unbalanced moments are, in Ο Β, -100, and in Ο G,- 50,
giving a total of -150 on the left of the joint; in Ο F, +
80, and in Ο D,
-200, giving a total of -120 on the right of the joint. The total unbalanced
moment at the joint, which is the difference between the total moment on the
left and on the right of the joint, is 30. This is now distributed in the respec-
tive propor~ions, as fol1ows:
/
924 ANALYSIS OF CONτiNUOUS FRAMES [Papcrs.

Το ΟΒ,
4
30 = 10
4+2+5+1
to OF,
2
30 = 5
4+2+5+1
to OD,
5
4+
2
+5+1
30 = 12.5
and, to Ο G,
1
4 + 2 + 5 + 1 30 = 2.5
The1·e is only one way to place the signs o:f the distributed moment~ so that
the total is the same on both sid€s o:f the joint. This is done by reducing the
excess negative moment on the le:ft and increasing the negative moment on
the right.
4.-At F, the unbalanced moment is + 60. The hinge has no sti:ffness. The
moment, then, is distributed between the member, F .0, and the hinge in the
ratio, 2 : Ο; all o:f it goes to the member. The total balanced moment is
+· 60- 60 =Ο, as it must be at a :free end.
5.-At G, the abutment is infinitely sti:ff and the unbalanced moment,
- 50, is distributed between the member, G Ο, and the abutment in the ratio,
1 : οο. The member gets none of it; the end stays fixed.
6.-At D, the unbalanced moment, -100, is distributed to D Ο and to D Ε
in the ratio o:f 5 : 3.
7.-At Ε, the unbalanced moment is -10 in the cantilever. Since the
cantilever has no sti:ffness, this unbalanced moment is distributed between the
beam, Ε D, and the cantilever in the ratio, 3 : Ο. This means that all o:f it
goes to Ε D.
All joints have now been balanced. Ν ext, carry over :from each end o:f
each member one-hal:f the distributed moment just written, reverse the sign,
and write it at the other end o:f the member. Thus, carry over, successively,
in Λ Β, Ο from Λ to Β and +·16.67 :from Β to Λ; in Β Ο,- 33.34 :from Β to . Ο
and -5.0 :from Ο to Β; in Ο F, +·
2.5 :from σ to F and +30 :from F to .0;
in Ο G, Ο :from G to Ο and -1.25 from σ to G; in Ο D, +
6.25 from Ο to D
and - 31.25 from D to Ο; and in D Ε, +18.75 :from D to Ε and + 5.00 :from
Ε to D.
Distribute the moments just caπied over exactly as the original fixed-end
moments were distributed. Thus, at Λ, +16.67 is distributed Ο to Λ Β (fixed-
ended) ; at Β, - 5.0 is distributed as - 1.67 and + 3.32; at Ο, the unbalanced
moment is (- 33.34 +· Ο) - C+ 30.00 - 31.25) = - 32.09 which is dis-
tributed as + 2.67, + 10.68, - 5.34, and - 13.35; at F, +
2.50 is distributed
as - 2.5 to the member; G is fixed-ended; at D, +
1.25 is distributed as- 0.78
and + 0.47; at Ε, the unbalanced + 18.75 is distributed to the member
as -18.75.
The moments distributed are now carried over as be:fore and then re-dis-
tributed; and the process is repeated as often as desired. The procedure should
:May, 1930.] ΔΝ ΔLYSIS OF CONTINUOUS FRΔ:M:ES 925

be stopped after each distribution, however, and a check made to see that
statics (.Σ Μ = Ο) is satisfied.
When it is felt that the process has gone far enough, all moments at each
end of each member are added to give the total moment at the joint. After
the moments at the joints have be~n determined, all other quantities, such as
moments and shears, may be obtained by applying the laws of statics.

CoNVERGENOE OF REsuLτs

The distribution herein has been carried out with more precision than is
ordinarily necessary, in order to show the convergence of the results. Το show
the rate of conveι·gence, the successive values of the moments at the joints
after successive distributions are given in Table 1.

TABLE 1.-CONVERGENOE OF RESULTS.

After one
Successive vιιlues distribution After two After thre.e After four After tlve After six
of bending di!itri bu tions distributions distri butionR distributions distributions
(two rows (four rows (sίχ rows (eig-ht rows (ten rows (twelve rows
moment at joint. of tlgures). of tlgures). of tlgures) . of tlguι·es). of tlgures). of tlgures).

.Α
Β
....... . ο
- 83.84
+ 16.67
-85.01
+ 17.50
- 36.79
+ 18 .39
- 36.97
+ 18.48
- 37.10
+ 18.!i5
- 87.18
ϊ~'ά'ji -90.00 - 112.66 -113.22 -114 .24 -114.23 -114.26
c f "CF + 75.00 + 99.66 +100.36 +101.82 +101 .36 +101.41
ι ...... ..
"CD -212 .50 - 257.10 -258 .09 -259.89 -259.88 -259.93
'' C G -47.50 - 44.83 - 44.55 - 44.36 - 44.31 - 44.28
D - 37.50 - 82 .03 -23.66 - 23.48 - 23.15 - 23.15
Ε - 10.00 - 10.00
F ······· ·.
....... ο
- 10.00
ο
-10.00
ο
- 10.00
ο ο
- 10.00
ο
G ........ - 50.00 - 51.25 - 52.59 - 52.73 -52.88 -52.86

For most purposes the computations might as well have been stopped
after the seoond distribution. Had this been done, the solution would have
appeared as shown in Fig. 2.
For any practical purpose the computation might in this case have been
stopped after the third distribution. In general, two or three distributions are
sufficient. This is not true in all instances, but in any case the exactness of
the solution at any stage will be indicated by the magnitude of the moments
carried over in the members.

V ARIATIONS OF ΤΗΕ METHOD

The writer has developed and used at di:fferent times several variations of
the method shown, but the original method is itself so simple and so easy to
remember that he finds himself inclined to discard the variants.
One variant is perhaps worth reco:iding. It is rather tedious to carry
moments out to the end o:f a member which is :free to rotate and then balance
the moment and carry it back again. This may be avoided by releasing the
free end once for all and leaving it :free. In this case, :for beams of constant
section, the sti:ffness of the beam is to be taken three-fourths as great* as
* The moment n eeded to produce a given rot ation at one end of a beam when the other
end is free is three-fourths as great as if the other end is tlxed .
926 ANALYSIS OF CONTINUOUS FRAMES [Papeι·s.

Ι
the relative --value would indicate. Afteι· tlιe end of tlιe beanι is once
L
released, no moments are carried over to it.
OoaaEoτiNG FOR SmE-SWAY
Single square or trapezoidal frames, portals, L-frames, box culverts, and
similar structures act as simple continuous beams if theι·e is no transverse
defl.ection. Ιί they are symmetrical as to form and loading, they will not defl.ect
sidewise and if they are restrained against sidewise movement, they cannot so
defl.ect.

®
-112.6
+ 10.7 C!C!C!"11":
- 33.3 ~"'~"' ~
+ 10.0 + ι+ ι +
Β -100.0 c
0,. . . ""
~ ,...;ο~ο -100.0 "'I"
:;: <"Jo<-Jg
"'0 -200.0
ι ι ι + 66.7 ι + + ι - 12.5
- 5.0 - 31.3
+ 3.j - 13.4
-35.0 -257.2

οοlό
"1"" Ιό
........
+ +
~
Α
FIG. 2.

Side-sway of frames due to dissymmetry of the frame is raι·ely an important

ίactor in design. Ooτrection :for side-sway may be made by a method which
may be applied also in cases of transverse loading on bents. The method is to
consider that the bent does not sway sidewise and analyze it as a series of con-
tinuous beams. The total shear in the legs will not now, except by accident,
equal the shear which is known to exist. The difference must be a force
which prevents side-sway.
Now, assume all joints held against rotation, but the top of theΘ moved
sidewise. Assume any series οί fixed-end moments in the legs such that alllegs
hav~ the same defl.ection. In this case for members of uniform section fixed-end
. Ι
nιoments 1n columns vary as L 2 • Distl'ibute these fixed-end ωoωents and find
the total shear in the legs. The changes in moments due to side-sway will then
be to the moments just computed in the same algebraic ratio as the total
unbalanced horizontal shear in the legs due to side-sway, \\·hen the frame is
analyzed as a continuous girder, is to the shear just computed.
~\1ay, 1930.1 ANΔLYSIS OF CONTINUOUS FRΔME 927

MuLη-SτoRIED BENTS

Bents οί rnoι·e than one story, subject to side-sway, either as a result οί

unbalanced loading or due to horizontal forces, may be solved by this method.
It is understood that exact solution of uch pι·oblems is not commonly of great
interest. It is the approximate e:ffect that is desired rather than exact analysis.
Το analyze by this method a two-story bent it will be necessary to make
two configurations-one for each story. From the assumed shear in each story
(producing, of cour e, shears in the other stories), a set of moment values may
be obtained. These may be combined to obtain the true sheaι·s, and fι·om the
true shears the trυe mornent ίol1ow.
GENERAL APPLICAτiON OF ΤΗΕ ΜΕΊ'ΗΟ])
The method herein indicated of distributing unbalanced moments may be
extended to include unbalaJιced joίnt foι·ce . As thus extended it has very
wide application. Horizontal or vertical rea.ctions may be distributed and
carried over and thus a quick estimate made of the e:ffect of many complicating
elements in design. The writer has used it in studying such problems as
continuous arch series, the e:ffect of the de:flection of suppω·tίng girders, and
other phenomena.
An obvious application of moment distribution occurs in the computation
of secondaι·y stresses in trusses. Many other applications wi]] doubtless sug-
gest themselves, but it has been thought best to restrict this paper chie:fly to
continuous frarnes in wh:ich the joints do not rnove.
00NCLUSION
The paper has been confined to a method of analysis, because it has seemed
wiser to so restrίct it. It is not then an oversig·ht that it does not deal with:
(1) Methods of constructing curves of maximum moments; (2) methods of
con~tructing curves of maximurn shears; ( 3) the importance of analyses for
continuity in the design of concι·ete girders; ( 4) :flexural stresses in concrete
columns; (5) methods of constructing in:fluence lίnes; (6) the degτee to which
continuity exists in ordinary steel frames; (7) continuity in welded steel
fι·ames; (8) plastic deformation beyond the yield point as an element in inter-
pι·eting secondary stress computations; (9) the e:ffect of time yie]d on moments
and shears in continuous concι·ete frames; (10) plastic flow of concrete as a
factor in the design of continuous concrete frames; (11) whether in concrete
frames it is better to g·uess at the moments, to take results from studies made
by Winkler fifty years ago, or to compute them; (12) the e:ffect of torsion of
connecting members; (13) the relative economy of continuous structures;
(14) the relative :flexibility of conti~uous structures; (15) the application of
methods of continuous frame analysis to the design of :flat s]abs; (16) proba-
ability of loading and reveι·sal of stress as factors in the design of continuous
ίι·ames; (17) the relation of precision in the determinat:ion of shears and /
moments to precision in the determination of fiber stresse , and a dozen other
considerations bearing on the design of continuous frames.
The writer has discussed several of these questions e1sewhere. He hopes
that readers will discuss some of them now.
928 ANALYSIS OF CONTINUOUS FRA::M:ES [Papers.

Α method of analysis has value if it is ultimately useful to the designer;

not otherwise. There are apparently three schools of thought as to the
value of analyses of continuous frames. Some say, "Since these problems
cannot be solved with exactness because of physical unceι·tainties, why try
to solve them at all ~" Others say, "The values of the moments and shears
cannot be found exactly ; do not try to find them exactly ; use a metl1od of
analysis which will combine reasonable precision with speed." Still όthers
say, ''It is best to be absolutely exact in the analysis and to introduce all
elements of judgment after making the analysis.''
The wl'iter belongs to the second school; he respects but finds di:fficulty in
understanding the viewpoint of the other two. Those who agree with his view-
point will find the method herein explained a useful guide to judgment in
design.
Members of the last named school of thought should note that the method
here presented is absolutely exact if absolute exactness is desired. It is a
method of successjve approxirnations; not an appr.oxirnate rnethod.